Since moving to memalloc_nofs_save/restore, SUNRPC has stopped setting the
GFP_NOIO flag on sk_allocation which the networking system uses to decide
when it is safe to use current->task_frag. The results of this are
unexpected corruption in task_frag when SUNRPC is involved in memory
reclaim.
The corruption can be seen in crashes, but the root cause is often
difficult to ascertain as a crashing machine's stack trace will have no
evidence of being near NFS or SUNRPC code. I believe this problem to
be much more pervasive than reports to the community may indicate.
Fix this by having kernel users of sockets that may corrupt task_frag due
to reclaim set sk_use_task_frag = false. Preemptively correcting this
situation for users that still set sk_allocation allows them to convert to
memalloc_nofs_save/restore without the same unexpected corruptions that are
sure to follow, unlikely to show up in testing, and difficult to bisect.
CC: Philipp Reisner <philipp.reisner@linbit.com>
CC: Lars Ellenberg <lars.ellenberg@linbit.com>
CC: "Christoph Böhmwalder" <christoph.boehmwalder@linbit.com>
CC: Jens Axboe <axboe@kernel.dk>
CC: Josef Bacik <josef@toxicpanda.com>
CC: Keith Busch <kbusch@kernel.org>
CC: Christoph Hellwig <hch@lst.de>
CC: Sagi Grimberg <sagi@grimberg.me>
CC: Lee Duncan <lduncan@suse.com>
CC: Chris Leech <cleech@redhat.com>
CC: Mike Christie <michael.christie@oracle.com>
CC: "James E.J. Bottomley" <jejb@linux.ibm.com>
CC: "Martin K. Petersen" <martin.petersen@oracle.com>
CC: Valentina Manea <valentina.manea.m@gmail.com>
CC: Shuah Khan <shuah@kernel.org>
CC: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
CC: David Howells <dhowells@redhat.com>
CC: Marc Dionne <marc.dionne@auristor.com>
CC: Steve French <sfrench@samba.org>
CC: Christine Caulfield <ccaulfie@redhat.com>
CC: David Teigland <teigland@redhat.com>
CC: Mark Fasheh <mark@fasheh.com>
CC: Joel Becker <jlbec@evilplan.org>
CC: Joseph Qi <joseph.qi@linux.alibaba.com>
CC: Eric Van Hensbergen <ericvh@gmail.com>
CC: Latchesar Ionkov <lucho@ionkov.net>
CC: Dominique Martinet <asmadeus@codewreck.org>
CC: Ilya Dryomov <idryomov@gmail.com>
CC: Xiubo Li <xiubli@redhat.com>
CC: Chuck Lever <chuck.lever@oracle.com>
CC: Jeff Layton <jlayton@kernel.org>
CC: Trond Myklebust <trond.myklebust@hammerspace.com>
CC: Anna Schumaker <anna@kernel.org>
CC: Steffen Klassert <steffen.klassert@secunet.com>
CC: Herbert Xu <herbert@gondor.apana.org.au>
Suggested-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: Benjamin Coddington <bcodding@redhat.com>
Reviewed-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Sockets that can be used while recursing into memory reclaim, like
those used by network block devices and file systems, mustn't use
current->task_frag: if the current process is already using it, then
the inner memory reclaim call would corrupt the task_frag structure.
To avoid this, sk_page_frag() uses ->sk_allocation to detect sockets
that mustn't use current->task_frag, assuming that those used during
memory reclaim had their allocation constraints reflected in
->sk_allocation.
This unfortunately doesn't cover all cases: in an attempt to remove all
usage of GFP_NOFS and GFP_NOIO, sunrpc stopped setting these flags in
->sk_allocation, and used memalloc_nofs critical sections instead.
This breaks the sk_page_frag() heuristic since the allocation
constraints are now stored in current->flags, which sk_page_frag()
can't read without risking triggering a cache miss and slowing down
TCP's fast path.
This patch creates a new field in struct sock, named sk_use_task_frag,
which sockets with memory reclaim constraints can set to false if they
can't safely use current->task_frag. In such cases, sk_page_frag() now
always returns the socket's page_frag (->sk_frag). The first user is
sunrpc, which needs to avoid using current->task_frag but can keep
->sk_allocation set to GFP_KERNEL otherwise.
Eventually, it might be possible to simplify sk_page_frag() by only
testing ->sk_use_task_frag and avoid relying on the ->sk_allocation
heuristic entirely (assuming other sockets will set ->sk_use_task_frag
according to their constraints in the future).
The new ->sk_use_task_frag field is placed in a hole in struct sock and
belongs to a cache line shared with ->sk_shutdown. Therefore it should
be hot and shouldn't have negative performance impacts on TCP's fast
path (sk_shutdown is tested just before the while() loop in
tcp_sendmsg_locked()).
Link: https://lore.kernel.org/netdev/b4d8cb09c913d3e34f853736f3f5628abfd7f4b6.1656699567.git.gnault@redhat.com/
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: Benjamin Coddington <bcodding@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The unregister check could be incorrectly triggered if a netdev
changes its type after register. That is possible for a tun device
using TUNSETLINK ioctl, resulting in mctp unregister failing
and the netdev unregister waiting forever.
This was encountered by https://github.com/openthread/openthread/issues/8523
Neither check at register or unregister is required. They were added in
an attempt to track down mctp_ptr being set unexpectedly, which should
not happen in normal operation.
Fixes: 7b1871af75 ("mctp: Warn if pointer is set for a wrong dev type")
Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
Link: https://lore.kernel.org/r/20221215054933.2403401-1-matt@codeconstruct.com.au
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
KSZ swithes used interrupts for detecting the phy link up and down.
During registering the interrupt handler, it used IRQF_TRIGGER_FALLING
flag. But this flag has to be retrieved from device tree instead of hard
coding in the driver, so removing the flag.
Fixes: ff319a6448 ("net: dsa: microchip: move interrupt handling logic from lan937x to ksz_common")
Reported-by: Christian Eggers <ceggers@arri.de>
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Link: https://lore.kernel.org/r/20221213101440.24667-1-arun.ramadoss@microchip.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
RFC1662 defines the start state for the crc16 FCS to be 0xffff, but
we're currently starting at zero.
This change uses the correct start state. We're only early in the
adoption for the serial binding, so there aren't yet any other users to
interface to.
Fixes: a0c2ccd9b5 ("mctp: Add MCTP-over-serial transport binding")
Reported-by: Harsh Tyagi <harshtya@google.com>
Tested-by: Harsh Tyagi <harshtya@google.com>
Signed-off-by: Jeremy Kerr <jk@codeconstruct.com.au>
Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The address of 32-bit extend capability is not qword aligned,
and may cause exception in some arch.
Fixes: 484963ce9f ("nfp: extend capability and control words")
Signed-off-by: Huanhuan Wang <huanhuan.wang@corigine.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Changheon Lee reported TCP socket leaks, with a nice repro.
It seems we leak TCP sockets with the following sequence:
1) SOF_TIMESTAMPING_TX_ACK is enabled on the socket.
Each ACK will cook an skb put in error queue, from __skb_tstamp_tx().
__skb_tstamp_tx() is using skb_clone(), unless
SOF_TIMESTAMPING_OPT_TSONLY was also requested.
2) If the application is also using MSG_ZEROCOPY, then we put in the
error queue cloned skbs that had a struct ubuf_info attached to them.
Whenever an struct ubuf_info is allocated, sock_zerocopy_alloc()
does a sock_hold().
As long as the cloned skbs are still in sk_error_queue,
socket refcount is kept elevated.
3) Application closes the socket, while error queue is not empty.
Since tcp_close() no longer purges the socket error queue,
we might end up with a TCP socket with at least one skb in
error queue keeping the socket alive forever.
This bug can be (ab)used to consume all kernel memory
and freeze the host.
We need to purge the error queue, with proper synchronization
against concurrent writers.
Fixes: 24bcbe1cc6 ("net: stream: don't purge sk_error_queue in sk_stream_kill_queues()")
Reported-by: Changheon Lee <darklight2357@icloud.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some memory allocated in myri10ge_probe_slices() is not released in the
error handling path of myri10ge_probe().
Add the corresponding kfree(), as already done in the remove function.
Fixes: 0dcffac1a3 ("myri10ge: add multislices support")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix the following smatch warning:
smatch warnings:
drivers/net/ethernet/microchip/vcap/vcap_api_debugfs.c:103 vcap_debugfs_show_rule_keyfield() error: uninitialized symbol 'value'.
drivers/net/ethernet/microchip/vcap/vcap_api_debugfs.c:106 vcap_debugfs_show_rule_keyfield() error: uninitialized symbol 'mask'.
In case the vcap field was VCAP_FIELD_U128 and the key was different
than IP6_S/DIP then the value and mask were not initialized, therefore
initialize them.
Fixes: 610c32b2ce ("net: microchip: vcap: Add vcap_get_rule")
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <error27@gmail.com>
Reviewed-by: Saeed Mahameed <saeed@kernel.org>
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
David Howells says:
====================
rxrpc: Fixes for I/O thread conversion/SACK table expansion
Here are some fixes for AF_RXRPC:
(1) Fix missing unlock in rxrpc's sendmsg.
(2) Fix (lack of) propagation of security settings to rxrpc_call.
(3) Fix NULL ptr deref in rxrpc_unuse_local().
(4) Fix problem with kthread_run() not invoking the I/O thread function if
the kthread gets stopped first. Possibly this should actually be
fixed in the kthread code.
(5) Fix locking problem as putting a peer (which may be done from RCU) may
now invoke kthread_stop().
(6) Fix switched parameters in a couple of trace calls.
(7) Fix I/O thread's checking for kthread stop to make sure it completes
all outstanding work before returning so that calls are cleaned up.
(8) Fix an uninitialised var in the new rxperf test server.
(9) Fix the return value of rxrpc_new_incoming_call() so that the checks
on it work correctly.
The patches fix at least one syzbot bug[1] and probably some others that
don't have reproducers[2][3][4]. I think it also fixes another[5], but
that showed another failure during testing that was different to the
original.
There's also an outstanding bug in rxrpc_put_peer()[6] that is fixed by a
combination of several patches in my rxrpc-next branch, but I haven't
included that here.
====================
Tested-by: Marc Dionne <marc.dionne@auristor.com>
Tested-by: kafs-testing+fedora36_64checkkafs-build-164@auristor.com
Signed-off-by: David S. Miller <davem@davemloft.net>
Dan Carpenter sayeth[1]:
The patch 5e6ef4f101: "rxrpc: Make the I/O thread take over the
call and local processor work" from Jan 23, 2020, leads to the
following Smatch static checker warning:
net/rxrpc/io_thread.c:283 rxrpc_input_packet()
warn: bool is not less than zero.
Fix this (for now) by changing rxrpc_new_incoming_call() to return an int
with 0 or error code rather than bool. Note that the actual return value
of rxrpc_input_packet() is currently ignored. I have a separate patch to
clean that up.
Fixes: 5e6ef4f101 ("rxrpc: Make the I/O thread take over the call and local processor work")
Reported-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
Link: http://lists.infradead.org/pipermail/linux-afs/2022-December/006123.html [1]
Signed-off-by: David S. Miller <davem@davemloft.net>
Dan Carpenter sayeth[1]:
The patch 75bfdbf2fc: "rxrpc: Implement an in-kernel rxperf server
for testing purposes" from Nov 3, 2022, leads to the following Smatch
static checker warning:
net/rxrpc/rxperf.c:337 rxperf_deliver_to_call()
error: uninitialized symbol 'ret'.
Fix this by initialising ret to 0. The value is only used for tracing
purposes in the rxperf server.
Fixes: 75bfdbf2fc ("rxrpc: Implement an in-kernel rxperf server for testing purposes")
Reported-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
Link: http://lists.infradead.org/pipermail/linux-afs/2022-December/006124.html [1]
Signed-off-by: David S. Miller <davem@davemloft.net>
The rxrpc I/O thread checks to see if there's any work it needs to do, and
if not, checks kthread_should_stop() before scheduling, and if it should
stop, breaks out of the loop and tries to clean up and exit.
This can, however, race with socket destruction, wherein outstanding calls
are aborted and released from the socket and then the socket unuses the
local endpoint, causing kthread_stop() to be issued. The abort is deferred
to the I/O thread and the event can by issued between the I/O thread
checking if there's any work to be done (such as processing call aborts)
and the stop being seen.
This results in the I/O thread stopping processing of events whilst call
cleanup events are still outstanding, leading to connections or other
objects still being around and uncleaned up, which can result in assertions
being triggered, e.g.:
rxrpc: AF_RXRPC: Leaked client conn 00000000e8009865 {2}
------------[ cut here ]------------
kernel BUG at net/rxrpc/conn_client.c:64!
Fix this by retrieving the kthread_should_stop() indication, then checking
to see if there's more work to do, and going back round the loop if there
is, and breaking out of the loop only if there wasn't.
This was triggered by a syzbot test that produced some other symptom[1].
Fixes: a275da62e8 ("rxrpc: Create a per-local endpoint receive queue and I/O thread")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
Link: https://lore.kernel.org/r/0000000000002b4a9f05ef2b616f@google.com/ [1]
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix the switched parameters on rxrpc_alloc_peer() and rxrpc_get_peer().
The ref argument and the why argument got mixed.
Fixes: 47c810a798 ("rxrpc: trace: Don't use __builtin_return_address for rxrpc_peer tracing")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that rxrpc_put_local() may call kthread_stop(), it can't be called
under spinlock as it might sleep. This can cause a problem in the peer
keepalive code in rxrpc as it tries to avoid dropping the peer_hash_lock
from the point it needs to re-add peer->keepalive_link to going round the
loop again in rxrpc_peer_keepalive_dispatch().
Fix this by just dropping the lock when we don't need it and accepting that
we'll have to take it again. This code is only called about every 20s for
each peer, so not very often.
This allows rxrpc_put_peer_unlocked() to be removed also.
If triggered, this bug produces an oops like the following, as reproduced
by a syzbot reproducer for a different oops[1]:
BUG: sleeping function called from invalid context at kernel/sched/completion.c:101
...
RCU nest depth: 0, expected: 0
3 locks held by kworker/u9:0/50:
#0: ffff88810e74a138 ((wq_completion)krxrpcd){+.+.}-{0:0}, at: process_one_work+0x294/0x636
#1: ffff8881013a7e20 ((work_completion)(&rxnet->peer_keepalive_work)){+.+.}-{0:0}, at: process_one_work+0x294/0x636
#2: ffff88817d366390 (&rxnet->peer_hash_lock){+.+.}-{2:2}, at: rxrpc_peer_keepalive_dispatch+0x2bd/0x35f
...
Call Trace:
<TASK>
dump_stack_lvl+0x4c/0x5f
__might_resched+0x2cf/0x2f2
__wait_for_common+0x87/0x1e8
kthread_stop+0x14d/0x255
rxrpc_peer_keepalive_dispatch+0x333/0x35f
rxrpc_peer_keepalive_worker+0x2e9/0x449
process_one_work+0x3c1/0x636
worker_thread+0x25f/0x359
kthread+0x1a6/0x1b5
ret_from_fork+0x1f/0x30
Fixes: a275da62e8 ("rxrpc: Create a per-local endpoint receive queue and I/O thread")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
Link: https://lore.kernel.org/r/0000000000002b4a9f05ef2b616f@google.com/ [1]
Signed-off-by: David S. Miller <davem@davemloft.net>
When starting a kthread, the __kthread_create_on_node() function, as called
from kthread_run(), waits for a completion to indicate that the task_struct
(or failure state) of the new kernel thread is available before continuing.
This does not wait, however, for the thread function to be invoked and,
indeed, will skip it if kthread_stop() gets called before it gets there.
If this happens, though, kthread_run() will have returned successfully,
indicating that the thread was started and returning the task_struct
pointer. The actual error indication is returned by kthread_stop().
Note that this is ambiguous, as the caller cannot tell whether the -EINTR
error code came from kthread() or from the thread function.
This was encountered in the new rxrpc I/O thread, where if the system is
being pounded hard by, say, syzbot, the check of KTHREAD_SHOULD_STOP can be
delayed long enough for kthread_stop() to get called when rxrpc releases a
socket - and this causes an oops because the I/O thread function doesn't
get started and thus doesn't remove the rxrpc_local struct from the
local_endpoints list.
Fix this by using a completion to wait for the thread to actually enter
rxrpc_io_thread(). This makes sure the thread can't be prematurely
stopped and makes sure the relied-upon cleanup is done.
Fixes: a275da62e8 ("rxrpc: Create a per-local endpoint receive queue and I/O thread")
Reported-by: syzbot+3538a6a72efa8b059c38@syzkaller.appspotmail.com
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: Hillf Danton <hdanton@sina.com>
Link: https://lore.kernel.org/r/000000000000229f1505ef2b6159@google.com/
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix the propagation of the security settings from sendmsg to the rxrpc_call
struct.
Fixes: f3441d4125 ("rxrpc: Copy client call parameters into rxrpc_call earlier")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
Signed-off-by: David S. Miller <davem@davemloft.net>
One of the error paths in rxrpc_do_sendmsg() doesn't unlock the call mutex
before returning. Fix it to do this.
Note that this still doesn't get rid of the checker warning:
../net/rxrpc/sendmsg.c:617:5: warning: context imbalance in 'rxrpc_do_sendmsg' - wrong count at exit
I think the interplay between the socket lock and the call's user_mutex may
be too complicated for checker to analyse, especially as
rxrpc_new_client_call_for_sendmsg(), which it calls, returns with the
call's user_mutex if successful but unconditionally drops the socket lock.
Fixes: e754eba685 ("rxrpc: Provide a cmsg to specify the amount of Tx data for a call")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
Signed-off-by: David S. Miller <davem@davemloft.net>
When TCF_EM_SIMPLE was introduced, it is supposed to be convenient
for ematch implementation:
https://lore.kernel.org/all/20050105110048.GO26856@postel.suug.ch/
"You don't have to, providing a 32bit data chunk without TCF_EM_SIMPLE
set will simply result in allocating & copy. It's an optimization,
nothing more."
So if an ematch module provides ops->datalen that means it wants a
complex data structure (saved in its em->data) instead of a simple u32
value. We should simply reject such a combination, otherwise this u32
could be misinterpreted as a pointer.
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Reported-and-tested-by: syzbot+4caeae4c7103813598ae@syzkaller.appspotmail.com
Reported-by: Jun Nie <jun.nie@linaro.org>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Tony Nguyen says:
====================
Intel Wired LAN Driver Updates 2022-12-15 (igc)
Muhammad Husaini Zulkifli says:
This patch series fixes bugs for the Time-Sensitive Networking(TSN)
Qbv Scheduling features.
An overview of each patch series is given below:
Patch 1: Using a first flag bit to schedule a packet to the next cycle if
packet cannot fit in current Qbv cycle.
Patch 2: Enable strict cycle for Qbv scheduling.
Patch 3: Prevent user to set basetime less than zero during tc config.
Patch 4: Allow the basetime enrollment with zero value.
Patch 5: Calculate the new end time value to exclude the time interval that
exceed the cycle time as user can specify the cycle time in tc config.
Patch 6: Resolve the HW bugs where the gate is not fully closed.
---
This contains the net patches from this original pull request:
https://lore.kernel.org/netdev/20221205212414.3197525-1-anthony.l.nguyen@intel.com/
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Extending the tail can have some unexpected side effects if a program uses
a helper like BPF_FUNC_skb_pull_data to read partial content beyond the
head skb headlen when all the skbs in the gso frag_list are linear with no
head_frag -
kernel BUG at net/core/skbuff.c:4219!
pc : skb_segment+0xcf4/0xd2c
lr : skb_segment+0x63c/0xd2c
Call trace:
skb_segment+0xcf4/0xd2c
__udp_gso_segment+0xa4/0x544
udp4_ufo_fragment+0x184/0x1c0
inet_gso_segment+0x16c/0x3a4
skb_mac_gso_segment+0xd4/0x1b0
__skb_gso_segment+0xcc/0x12c
udp_rcv_segment+0x54/0x16c
udp_queue_rcv_skb+0x78/0x144
udp_unicast_rcv_skb+0x8c/0xa4
__udp4_lib_rcv+0x490/0x68c
udp_rcv+0x20/0x30
ip_protocol_deliver_rcu+0x1b0/0x33c
ip_local_deliver+0xd8/0x1f0
ip_rcv+0x98/0x1a4
deliver_ptype_list_skb+0x98/0x1ec
__netif_receive_skb_core+0x978/0xc60
Fix this by marking these skbs as GSO_DODGY so segmentation can handle
the tail updates accordingly.
Fixes: 3dcbdb134f ("net: gso: Fix skb_segment splat when splitting gso_size mangled skb having linear-headed frag_list")
Signed-off-by: Sean Tranchetti <quic_stranche@quicinc.com>
Signed-off-by: Subash Abhinov Kasiviswanathan <quic_subashab@quicinc.com>
Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>
Link: https://lore.kernel.org/r/1671084718-24796-1-git-send-email-quic_subashab@quicinc.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Take the instance lock around devlink_nl_fill() when dumping,
doit takes it already.
We are only dumping basic info so in the worst case we were risking
data races around the reload statistics. Until the big devlink mutex
was removed all relevant code was protected by it, so the missing
instance lock was not exposed.
Fixes: d3efc2a6a6 ("net: devlink: remove devlink_mutex")
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://lore.kernel.org/r/20221216044122.1863550-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The #ifdef check is incorrect and leads to a warning:
drivers/net/ethernet/ti/am65-cpsw-nuss.c:1679:13: error: 'am65_cpsw_nuss_remove_rx_chns' defined but not used [-Werror=unused-function]
1679 | static void am65_cpsw_nuss_remove_rx_chns(void *data)
It's better to remove the #ifdef here and use the modern
SYSTEM_SLEEP_PM_OPS() macro instead.
Fixes: 24bc19b05f ("net: ethernet: ti: am65-cpsw: Add suspend/resume support")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>
Link: https://lore.kernel.org/r/20221215163918.611609-1-arnd@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
-----BEGIN PGP SIGNATURE-----
iHUEABYIAB0WIQTFp0I1jqZrAX+hPRXbK58LschIgwUCY5yc9AAKCRDbK58LschI
g3n4AP4heagqBHPH+WcC0N/Vc4K2dmPmil4ZTuZ/Xt+EMrX0MQEAygWsa272V2C9
vx51DOu5/D+DPC20/1+mRpEnC4JIqAA=
=utn+
-----END PGP SIGNATURE-----
Merge tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
Daniel Borkmann says:
====================
pull-request: bpf 2022-12-16
We've added 7 non-merge commits during the last 2 day(s) which contain
a total of 9 files changed, 119 insertions(+), 36 deletions(-).
1) Fix for recent syzkaller XDP dispatcher update splat, from Jiri Olsa.
2) Fix BPF program refcount leak in LSM attachment failure path,
from Milan Landaverde.
3) Fix BPF program type in map compatibility check for fext,
from Toke Høiland-Jørgensen.
4) Fix a BPF selftest compilation error under !CONFIG_SMP config,
from Yonghong Song.
5) Fix CI to enable CONFIG_FUNCTION_ERROR_INJECTION after it got changed
to a prompt, from Song Liu.
6) Various BPF documentation fixes for socket local storage,
from Donald Hunter.
* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
selftests/bpf: Add a test for using a cpumap from an freplace-to-XDP program
bpf: Resolve fext program type when checking map compatibility
bpf: Synchronize dispatcher update with bpf_dispatcher_xdp_func
bpf: prevent leak of lsm program after failed attach
selftests/bpf: Select CONFIG_FUNCTION_ERROR_INJECTION
selftests/bpf: Fix a selftest compilation error with CONFIG_SMP=n
docs/bpf: Reword docs for BPF_MAP_TYPE_SK_STORAGE
====================
Link: https://lore.kernel.org/r/20221216174540.16598-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The commit mentioned below causes the ovs_flow_tbl_lookup() function
to be called with the masked key. However, it's supposed to be called
with the unmasked key. This due to the fact that the datapath supports
installing wider flows, and OVS relies on this behavior. For example
if ipv4(src=1.1.1.1/192.0.0.0, dst=1.1.1.2/192.0.0.0) exists, a wider
flow (smaller mask) of ipv4(src=192.1.1.1/128.0.0.0,dst=192.1.1.2/
128.0.0.0) is allowed to be added.
However, if we try to add a wildcard rule, the installation fails:
$ ovs-appctl dpctl/add-flow system@myDP "in_port(1),eth_type(0x0800), \
ipv4(src=1.1.1.1/192.0.0.0,dst=1.1.1.2/192.0.0.0,frag=no)" 2
$ ovs-appctl dpctl/add-flow system@myDP "in_port(1),eth_type(0x0800), \
ipv4(src=192.1.1.1/0.0.0.0,dst=49.1.1.2/0.0.0.0,frag=no)" 2
ovs-vswitchd: updating flow table (File exists)
The reason is that the key used to determine if the flow is already
present in the system uses the original key ANDed with the mask.
This results in the IP address not being part of the (miniflow) key,
i.e., being substituted with an all-zero value. When doing the actual
lookup, this results in the key wrongfully matching the first flow,
and therefore the flow does not get installed.
This change reverses the commit below, but rather than having the key
on the stack, it's allocated.
Fixes: 190aa3e778 ("openvswitch: Fix Frame-size larger than 1024 bytes warning.")
Signed-off-by: Eelco Chaudron <echaudro@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jakub Kicinski says:
====================
devlink: region snapshot locking fix and selftest adjustments
Minor fix for region snapshot locking and adjustments to selftests.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
NetworkManager (and other daemons) may bring the interface up
and cause failures in quiescence checks. Print a helpful warning,
and take the interface down again.
I seem to forget about this every time I run these tests on a new VM.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
$number + > bash means redirect FD $number, e.g. commonly
used 2> redirects stderr (fd 2). The test uses 8192> to
write the number 8192 to a file, this results in:
./devlink.sh: line 499: 8192: Bad file descriptor
Oddly the test also papers over this issue by checking
for failure (expecting an error rather than success)
so it passes, anyway.
Fixes: ff18176ad8 ("selftests: Add a test of large binary to devlink health test")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Netdevsim triggers a splat on reload, when it destroys regions
with snapshots pending:
WARNING: CPU: 1 PID: 787 at net/core/devlink.c:6291 devlink_region_snapshot_del+0x12e/0x140
CPU: 1 PID: 787 Comm: devlink Not tainted 6.1.0-07460-g7ae9888d6e1c #580
RIP: 0010:devlink_region_snapshot_del+0x12e/0x140
Call Trace:
<TASK>
devl_region_destroy+0x70/0x140
nsim_dev_reload_down+0x2f/0x60 [netdevsim]
devlink_reload+0x1f7/0x360
devlink_nl_cmd_reload+0x6ce/0x860
genl_family_rcv_msg_doit.isra.0+0x145/0x1c0
This is the locking assert in devlink_region_snapshot_del(),
we're supposed to be holding the region->snapshot_lock here.
Fixes: 2dec18ad82 ("net: devlink: remove region snapshots list dependency on devlink->lock")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Russell King correctly pointed out that the MAC_2500FD capability is
already added for port 5 (if not in RGMII mode) and port 6 (which only
supports SGMII) by mt7531_mac_port_get_caps. Remove the reduntant
setting of this capability flag which was added by a previous commit.
Fixes: e19de30d20 ("net: dsa: mt7530: add support for in-band link status")
Reported-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Link: https://lore.kernel.org/r/Y5qY7x6la5TxZxzX@makrotopia.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The default setting of end_time minus start_time is whole 1 second.
Thus, if it's not being configured in any GCL entry then it will be
staying at original 1 second.
This patch is changing the start_time and end_time to be end_time as
if setting zero will be having weird HW behavior where the gate will
not be fully closed.
Fixes: ec50a9d437 ("igc: Add support for taprio offloading")
Signed-off-by: Tan Tee Min <tee.min.tan@linux.intel.com>
Signed-off-by: Muhammad Husaini Zulkifli <muhammad.husaini.zulkifli@intel.com>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Qbv users can specify a cycle time that is not equal to the total GCL
intervals. Hence, recalculation is necessary here to exclude the time
interval that exceeds the cycle time. As those GCL which exceeds the
cycle time will be truncated.
According to IEEE Std. 802.1Q-2018 section 8.6.9.2, once the end of
the list is reached, it will switch to the END_OF_CYCLE state and
leave the gates in the same state until the next cycle is started.
Fixes: ec50a9d437 ("igc: Add support for taprio offloading")
Signed-off-by: Tan Tee Min <tee.min.tan@linux.intel.com>
Signed-off-by: Muhammad Husaini Zulkifli <muhammad.husaini.zulkifli@intel.com>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Introduce qbv_enable flag in igc_adapter struct to store the Qbv on/off.
So this allow the BaseTime to enroll with zero value.
Fixes: 61572d5f8f ("igc: Simplify TSN flags handling")
Signed-off-by: Muhammad Husaini Zulkifli <muhammad.husaini.zulkifli@intel.com>
Signed-off-by: Tan Tee Min <tee.min.tan@linux.intel.com>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Using the tc qdisc command, the user can set basetime to any value.
Checking should be done on the driver's side to prevent registering
basetime values that are less than zero.
Fixes: ec50a9d437 ("igc: Add support for taprio offloading")
Signed-off-by: Muhammad Husaini Zulkifli <muhammad.husaini.zulkifli@intel.com>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Configuring strict cycle mode in the controller forces more well
behaved transmissions when taprio is offloaded.
When set this strict_cycle and strict_end, transmission is not
enabled if the whole packet cannot be completed before end of
the Qbv cycle.
Fixes: 82faa9b799 ("igc: Add support for ETF offloading")
Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: Aravindhan Gunasekaran <aravindhan.gunasekaran@intel.com>
Signed-off-by: Muhammad Husaini Zulkifli <muhammad.husaini.zulkifli@intel.com>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
The I225 hardware has a limitation that packets can only be scheduled
in the [0, cycle-time] interval. So, scheduling a packet to the start
of the next cycle doesn't usually work.
To overcome this, we use the Transmit Descriptor first flag to indicates
that a packet should be the first packet (from a queue) in a cycle
according to the section 7.5.2.9.3.4 The First Packet on Each QBV Cycle
in Intel Discrete I225/6 User Manual.
But this only works if there was any packet from that queue during the
current cycle, to avoid this issue, we issue an empty packet if that's
not the case. Also require one more descriptor to be available, to take
into account the empty packet that might be issued.
Test Setup:
Talker: Use l2_tai to generate the launchtime into packet load.
Listener: Use timedump.c to compute the delta between packet arrival
and LaunchTime packet payload.
Test Result:
Before:
1666000610127300000,1666000610127300096,96,621273
1666000610127400000,1666000610127400192,192,621274
1666000610127500000,1666000610127500032,32,621275
1666000610127600000,1666000610127600128,128,621276
1666000610127700000,1666000610127700224,224,621277
1666000610127800000,1666000610127800064,64,621278
1666000610127900000,1666000610127900160,160,621279
1666000610128000000,1666000610128000000,0,621280
1666000610128100000,1666000610128100096,96,621281
1666000610128200000,1666000610128200192,192,621282
1666000610128300000,1666000610128300032,32,621283
1666000610128400000,1666000610128301056,-98944,621284
1666000610128500000,1666000610128302080,-197920,621285
1666000610128600000,1666000610128302848,-297152,621286
1666000610128700000,1666000610128303872,-396128,621287
1666000610128800000,1666000610128304896,-495104,621288
1666000610128900000,1666000610128305664,-594336,621289
1666000610129000000,1666000610128306688,-693312,621290
1666000610129100000,1666000610128307712,-792288,621291
1666000610129200000,1666000610128308480,-891520,621292
1666000610129300000,1666000610128309504,-990496,621293
1666000610129400000,1666000610128310528,-1089472,621294
1666000610129500000,1666000610128311296,-1188704,621295
1666000610129600000,1666000610128312320,-1287680,621296
1666000610129700000,1666000610128313344,-1386656,621297
1666000610129800000,1666000610128314112,-1485888,621298
1666000610129900000,1666000610128315136,-1584864,621299
1666000610130000000,1666000610128316160,-1683840,621300
1666000610130100000,1666000610128316928,-1783072,621301
1666000610130200000,1666000610128317952,-1882048,621302
1666000610130300000,1666000610128318976,-1981024,621303
1666000610130400000,1666000610128319744,-2080256,621304
1666000610130500000,1666000610128320768,-2179232,621305
1666000610130600000,1666000610128321792,-2278208,621306
1666000610130700000,1666000610128322816,-2377184,621307
1666000610130800000,1666000610128323584,-2476416,621308
1666000610130900000,1666000610128324608,-2575392,621309
1666000610131000000,1666000610128325632,-2674368,621310
1666000610131100000,1666000610128326400,-2773600,621311
1666000610131200000,1666000610128327424,-2872576,621312
1666000610131300000,1666000610128328448,-2971552,621313
1666000610131400000,1666000610128329216,-3070784,621314
1666000610131500000,1666000610131500032,32,621315
1666000610131600000,1666000610131600128,128,621316
1666000610131700000,1666000610131700224,224,621317
After:
1666073510646200000,1666073510646200064,64,2676462
1666073510646300000,1666073510646300160,160,2676463
1666073510646400000,1666073510646400256,256,2676464
1666073510646500000,1666073510646500096,96,2676465
1666073510646600000,1666073510646600192,192,2676466
1666073510646700000,1666073510646700032,32,2676467
1666073510646800000,1666073510646800128,128,2676468
1666073510646900000,1666073510646900224,224,2676469
1666073510647000000,1666073510647000064,64,2676470
1666073510647100000,1666073510647100160,160,2676471
1666073510647200000,1666073510647200256,256,2676472
1666073510647300000,1666073510647300096,96,2676473
1666073510647400000,1666073510647400192,192,2676474
1666073510647500000,1666073510647500032,32,2676475
1666073510647600000,1666073510647600128,128,2676476
1666073510647700000,1666073510647700224,224,2676477
1666073510647800000,1666073510647800064,64,2676478
1666073510647900000,1666073510647900160,160,2676479
1666073510648000000,1666073510648000000,0,2676480
1666073510648100000,1666073510648100096,96,2676481
1666073510648200000,1666073510648200192,192,2676482
1666073510648300000,1666073510648300032,32,2676483
1666073510648400000,1666073510648400128,128,2676484
1666073510648500000,1666073510648500224,224,2676485
1666073510648600000,1666073510648600064,64,2676486
1666073510648700000,1666073510648700160,160,2676487
1666073510648800000,1666073510648800000,0,2676488
1666073510648900000,1666073510648900096,96,2676489
1666073510649000000,1666073510649000192,192,2676490
1666073510649100000,1666073510649100032,32,2676491
1666073510649200000,1666073510649200128,128,2676492
1666073510649300000,1666073510649300224,224,2676493
1666073510649400000,1666073510649400064,64,2676494
1666073510649500000,1666073510649500160,160,2676495
1666073510649600000,1666073510649600000,0,2676496
1666073510649700000,1666073510649700096,96,2676497
1666073510649800000,1666073510649800192,192,2676498
1666073510649900000,1666073510649900032,32,2676499
1666073510650000000,1666073510650000128,128,2676500
Fixes: 82faa9b799 ("igc: Add support for ETF offloading")
Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Co-developed-by: Aravindhan Gunasekaran <aravindhan.gunasekaran@intel.com>
Signed-off-by: Aravindhan Gunasekaran <aravindhan.gunasekaran@intel.com>
Co-developed-by: Muhammad Husaini Zulkifli <muhammad.husaini.zulkifli@intel.com>
Signed-off-by: Muhammad Husaini Zulkifli <muhammad.husaini.zulkifli@intel.com>
Signed-off-by: Malli C <mallikarjuna.chilakala@intel.com>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
In the blamed commit, it was not noticed that one implementation of
chip->info->ops->phylink_get_caps(), called by mv88e6xxx_get_caps(),
may access hardware registers, and in doing so, it takes the
mv88e6xxx_reg_lock(). Namely, this is mv88e6352_phylink_get_caps().
This is a problem because mv88e6xxx_get_caps(), apart from being
a top-level function (method invoked by dsa_switch_ops), is now also
directly called from mv88e6xxx_setup_port(), which runs under the
mv88e6xxx_reg_lock() taken by mv88e6xxx_setup(). Therefore, when running
on mv88e6352, the reg_lock would be acquired a second time and the
system would deadlock on driver probe.
The things that mv88e6xxx_setup() can compete with in terms of register
access with are the IRQ handlers and MDIO bus operations registered by
mv88e6xxx_probe(). So there is a real need to acquire the register lock.
The register lock can, in principle, be dropped and re-acquired pretty
much at will within the driver, as long as no operations that involve
waiting for indirect access to complete (essentially, callers of
mv88e6xxx_smi_direct_wait() and mv88e6xxx_wait_mask()) are interrupted
with the lock released. However, I would guess that in mv88e6xxx_setup(),
the critical section is kept open for such a long time just in order to
optimize away multiple lock/unlock operations on the registers.
We could, in principle, drop the reg_lock right before the
mv88e6xxx_setup_port() -> mv88e6xxx_get_caps() call, and
re-acquire it immediately afterwards. But this would look ugly, because
mv88e6xxx_setup_port() would release a lock which it didn't acquire, but
the caller did.
A cleaner solution to this issue comes from the observation that struct
mv88e6xxxx_ops methods generally assume they are called with the
reg_lock already acquired. Whereas mv88e6352_phylink_get_caps() is more
the exception rather than the norm, in that it acquires the lock itself.
Let's enforce the same locking pattern/convention for
chip->info->ops->phylink_get_caps() as well, and make
mv88e6xxx_get_caps(), the top-level function, acquire the register lock
explicitly, for this one implementation that will access registers for
port 4 to work properly.
This means that mv88e6xxx_setup_port() will no longer call the top-level
function, but the low-level mv88e6xxx_ops method which expects the
correct calling context (register lock held).
Compared to chip->info->ops->phylink_get_caps(), mv88e6xxx_get_caps()
also fixes up the supported_interfaces bitmap for internal ports, since
that can be done generically and does not require per-switch knowledge.
That's code which will no longer execute, however mv88e6xxx_setup_port()
doesn't need that. It just needs to look at the mac_capabilities bitmap.
Fixes: cc1049ccee ("net: dsa: mv88e6xxx: fix speed setting for CPU/DSA ports")
Reported-by: Maksim Kiselev <bigunclemax@gmail.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Tested-by: Maksim Kiselev <bigunclemax@gmail.com>
Link: https://lore.kernel.org/r/20221214110120.3368472-1-vladimir.oltean@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
This patch fixes the error "ravb 11c20000.ethernet eth0: failed to switch
device to config mode" during unbind.
We are doing register access after pm_runtime_put_sync().
We usually do cleanup in reverse order of init. Currently in
remove(), the "pm_runtime_put_sync" is not in reverse order.
Probe
reset_control_deassert(rstc);
pm_runtime_enable(&pdev->dev);
pm_runtime_get_sync(&pdev->dev);
remove
pm_runtime_put_sync(&pdev->dev);
unregister_netdev(ndev);
..
ravb_mdio_release(priv);
pm_runtime_disable(&pdev->dev);
Consider the call to unregister_netdev()
unregister_netdev->unregister_netdevice_queue->rollback_registered_many
that calls the below functions which access the registers after
pm_runtime_put_sync()
1) ravb_get_stats
2) ravb_close
Fixes: c156633f13 ("Renesas Ethernet AVB driver proper")
Cc: stable@vger.kernel.org
Signed-off-by: Biju Das <biju.das.jz@bp.renesas.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Link: https://lore.kernel.org/r/20221214105118.2495313-1-biju.das.jz@bp.renesas.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
We should set the return value to -ENOMEM explicitly when
create_singlethread_workqueue() fails in stmmac_dvr_probe(),
otherwise we'll lose the error value.
Fixes: a137f3f27f ("net: stmmac: fix possible memory leak in stmmac_dvr_probe()")
Signed-off-by: Gaosheng Cui <cuigaosheng1@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Link: https://lore.kernel.org/r/20221214080117.3514615-1-cuigaosheng1@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
There is a memory leaks reported by kmemleak:
unreferenced object 0xffff888116111000 (size 2048):
comm "modprobe", pid 817, jiffies 4294759745 (age 76.502s)
hex dump (first 32 bytes):
00 c4 0a 04 81 88 ff ff 08 10 11 16 81 88 ff ff ................
08 10 11 16 81 88 ff ff 00 00 00 00 00 00 00 00 ................
backtrace:
[<ffffffff815bcd82>] kmalloc_trace+0x22/0x60
[<ffffffff827e20ee>] phy_device_create+0x4e/0x90
[<ffffffff827e6072>] get_phy_device+0xd2/0x220
[<ffffffff827e7844>] mdiobus_scan+0xa4/0x2e0
[<ffffffff827e8be2>] __mdiobus_register+0x482/0x8b0
[<ffffffffa01f5d24>] r6040_init_one+0x714/0xd2c [r6040]
...
The problem occurs in probe process as follows:
r6040_init_one:
mdiobus_register
mdiobus_scan <- alloc and register phy_device,
the reference count of phy_device is 3
r6040_mii_probe
phy_connect <- connect to the first phy_device,
so the reference count of the first
phy_device is 4, others are 3
register_netdev <- fault inject succeeded, goto error handling path
// error handling path
err_out_mdio_unregister:
mdiobus_unregister(lp->mii_bus);
err_out_mdio:
mdiobus_free(lp->mii_bus); <- the reference count of the first
phy_device is 1, it is not released
and other phy_devices are released
// similarly, the remove process also has the same problem
The root cause is traced to the phy_device is not disconnected when
removes one r6040 device in r6040_remove_one() or on error handling path
after r6040_mii probed successfully. In r6040_mii_probe(), a net ethernet
device is connected to the first PHY device of mii_bus, in order to
notify the connected driver when the link status changes, which is the
default behavior of the PHY infrastructure to handle everything.
Therefore the phy_device should be disconnected when removes one r6040
device or on error handling path.
Fix it by adding phy_disconnect() when removes one r6040 device or on
error handling path after r6040_mii probed successfully.
Fixes: 3831861b4a ("r6040: implement phylib")
Signed-off-by: Li Zetao <lizetao1@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Link: https://lore.kernel.org/r/20221213125614.927754-1-lizetao1@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
There is a race resulting in alive SOCK_SEQPACKET socket
may change its state from TCP_ESTABLISHED to TCP_CLOSE:
unix_release_sock(peer) unix_dgram_sendmsg(sk)
sock_orphan(peer)
sock_set_flag(peer, SOCK_DEAD)
sock_alloc_send_pskb()
if !(sk->sk_shutdown & SEND_SHUTDOWN)
OK
if sock_flag(peer, SOCK_DEAD)
sk->sk_state = TCP_CLOSE
sk->sk_shutdown = SHUTDOWN_MASK
After that socket sk remains almost normal: it is able to connect, listen, accept
and recvmsg, while it can't sendmsg.
Since this is the only possibility for alive SOCK_SEQPACKET to change
the state in such way, we should better fix this strange and potentially
danger corner case.
Note, that we will return EPIPE here like this is normally done in sock_alloc_send_pskb().
Originally used ECONNREFUSED looks strange, since it's strange to return
a specific retval in dependence of race in kernel, when user can't affect on this.
Also, move TCP_CLOSE assignment for SOCK_DGRAM sockets under state lock
to fix race with unix_dgram_connect():
unix_dgram_connect(other) unix_dgram_sendmsg(sk)
unix_peer(sk) = NULL
unix_state_unlock(sk)
unix_state_double_lock(sk, other)
sk->sk_state = TCP_ESTABLISHED
unix_peer(sk) = other
unix_state_double_unlock(sk, other)
sk->sk_state = TCP_CLOSED
This patch fixes both of these races.
Fixes: 83301b5367 ("af_unix: Set TCP_ESTABLISHED for datagram sockets too")
Signed-off-by: Kirill Tkhai <tkhai@ya.ru>
Link: https://lore.kernel.org/r/135fda25-22d5-837a-782b-ceee50e19844@ya.ru
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
This adds a simple test for inserting an XDP program into a cpumap that is
"owned" by an XDP program that was loaded as PROG_TYPE_EXT (as libxdp
does). Prior to the kernel fix this would fail because the map type
ownership would be set to PROG_TYPE_EXT instead of being resolved to
PROG_TYPE_XDP.
v5:
- Fix a few nits from Andrii, add his ACK
v4:
- Use skeletons for selftest
v3:
- Update comment to better explain the cause
- Add Yonghong's ACK
Acked-by: Yonghong Song <yhs@fb.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://lore.kernel.org/r/20221214230254.790066-2-toke@redhat.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
The bpf_prog_map_compatible() check makes sure that BPF program types are
not mixed inside BPF map types that can contain programs (tail call maps,
cpumaps and devmaps). It does this by setting the fields of the map->owner
struct to the values of the first program being checked against, and
rejecting any subsequent programs if the values don't match.
One of the values being set in the map owner struct is the program type,
and since the code did not resolve the prog type for fext programs, the map
owner type would be set to PROG_TYPE_EXT and subsequent loading of programs
of the target type into the map would fail.
This bug is seen in particular for XDP programs that are loaded as
PROG_TYPE_EXT using libxdp; these cannot insert programs into devmaps and
cpumaps because the check fails as described above.
Fix the bug by resolving the fext program type to its target program type
as elsewhere in the verifier.
v3:
- Add Yonghong's ACK
Fixes: f45d5b6ce2 ("bpf: generalise tail call map compatibility check")
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://lore.kernel.org/r/20221214230254.790066-1-toke@redhat.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Fix a slab-out-of-bounds read that occurs in nla_put() called from
nfc_genl_send_target() when target->sensb_res_len, which is duplicated
from an nfc_target in pn533, is too large as the nfc_target is not
properly initialized and retains garbage values. Clear nfc_targets with
memset() before they are used.
Found by a modified version of syzkaller.
BUG: KASAN: slab-out-of-bounds in nla_put
Call Trace:
memcpy
nla_put
nfc_genl_dump_targets
genl_lock_dumpit
netlink_dump
__netlink_dump_start
genl_family_rcv_msg_dumpit
genl_rcv_msg
netlink_rcv_skb
genl_rcv
netlink_unicast
netlink_sendmsg
sock_sendmsg
____sys_sendmsg
___sys_sendmsg
__sys_sendmsg
do_syscall_64
Fixes: 673088fb42 ("NFC: pn533: Send ATR_REQ directly for active device detection")
Fixes: 361f3cb7f9 ("NFC: DEP link hook implementation for pn533")
Signed-off-by: Minsuk Kang <linuxlovemin@yonsei.ac.kr>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20221214015139.119673-1-linuxlovemin@yonsei.ac.kr
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Before enetc_clean_rx_ring_xdp() calls xdp_do_redirect(), each software
BD in the RX ring between index orig_i and i can have one of 2 refcount
values on its page.
We are the owner of the current buffer that is being processed, so the
refcount will be at least 1.
If the current owner of the buffer at the diametrically opposed index
in the RX ring (i.o.w, the other half of this page) has not yet called
kfree(), this page's refcount could even be 2.
enetc_page_reusable() in enetc_flip_rx_buff() tests for the page
refcount against 1, and [ if it's 2 ] does not attempt to reuse it.
But if enetc_flip_rx_buff() is put after the xdp_do_redirect() call,
the page refcount can have one of 3 values. It can also be 0, if there
is no owner of the other page half, and xdp_do_redirect() for this
buffer ran so far that it triggered a flush of the devmap/cpumap bulk
queue, and the consumers of those bulk queues also freed the buffer,
all by the time xdp_do_redirect() returns the execution back to enetc.
This is the reason why enetc_flip_rx_buff() is called before
xdp_do_redirect(), but there is a big flaw with that reasoning:
enetc_flip_rx_buff() will set rx_swbd->page = NULL on both sides of the
enetc_page_reusable() branch, and if xdp_do_redirect() returns an error,
we call enetc_xdp_free(), which does not deal gracefully with that.
In fact, what happens is quite special. The page refcounts start as 1.
enetc_flip_rx_buff() figures they're reusable, transfers these
rx_swbd->page pointers to a different rx_swbd in enetc_reuse_page(), and
bumps the refcount to 2. When xdp_do_redirect() later returns an error,
we call the no-op enetc_xdp_free(), but we still haven't lost the
reference to that page. A copy of it is still at rx_ring->next_to_alloc,
but that has refcount 2 (and there are no concurrent owners of it in
flight, to drop the refcount). What really kills the system is when
we'll flip the rx_swbd->page the second time around. With an updated
refcount of 2, the page will not be reusable and we'll really leak it.
Then enetc_new_page() will have to allocate more pages, which will then
eventually leak again on further errors from xdp_do_redirect().
The problem, summarized, is that we zeroize rx_swbd->page before we're
completely done with it, and this makes it impossible for the error path
to do something with it.
Since the packet is potentially multi-buffer and therefore the
rx_swbd->page is potentially an array, manual passing of the old
pointers between enetc_flip_rx_buff() and enetc_xdp_free() is a bit
difficult.
For the sake of going with a simple solution, we accept the possibility
of racing with xdp_do_redirect(), and we move the flip procedure to
execute only on the redirect success path. By racing, I mean that the
page may be deemed as not reusable by enetc (having a refcount of 0),
but there will be no leak in that case, either.
Once we accept that, we have something better to do with buffers on
XDP_REDIRECT failure. Since we haven't performed half-page flipping yet,
we won't, either (and this way, we can avoid enetc_xdp_free()
completely, which gives the entire page to the slab allocator).
Instead, we'll call enetc_xdp_drop(), which will recycle this half of
the buffer back to the RX ring.
Fixes: 9d2b68cc10 ("net: enetc: add support for XDP_REDIRECT")
Suggested-by: Lorenzo Bianconi <lorenzo.bianconi@redhat.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://lore.kernel.org/r/20221213001908.2347046-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Hao Sun reported crash in dispatcher image [1].
Currently we don't have any sync between bpf_dispatcher_update and
bpf_dispatcher_xdp_func, so following race is possible:
cpu 0: cpu 1:
bpf_prog_run_xdp
...
bpf_dispatcher_xdp_func
in image at offset 0x0
bpf_dispatcher_update
update image at offset 0x800
bpf_dispatcher_update
update image at offset 0x0
in image at offset 0x0 -> crash
Fixing this by synchronizing dispatcher image update (which is done
in bpf_dispatcher_update function) with bpf_dispatcher_xdp_func that
reads and execute the dispatcher image.
Calling synchronize_rcu after updating and installing new image ensures
that readers leave old image before it's changed in the next dispatcher
update. The update itself is locked with dispatcher's mutex.
The bpf_prog_run_xdp is called under local_bh_disable and synchronize_rcu
will wait for it to leave [2].
[1] https://lore.kernel.org/bpf/Y5SFho7ZYXr9ifRn@krava/T/#m00c29ece654bc9f332a17df493bbca33e702896c
[2] https://lore.kernel.org/bpf/0B62D35A-E695-4B7A-A0D4-774767544C1A@gmail.com/T/#mff43e2c003ae99f4a38f353c7969be4c7162e877
Reported-by: Hao Sun <sunhao.th@gmail.com>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Acked-by: Yonghong Song <yhs@fb.com>
Acked-by: Paul E. McKenney <paulmck@kernel.org>
Link: https://lore.kernel.org/r/20221214123542.1389719-1-jolsa@kernel.org
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
In [0], we added the ability to bpf_prog_attach LSM programs to cgroups,
but in our validation to make sure the prog is meant to be attached to
BPF_LSM_CGROUP, we return too early if the check fails. This results in
lack of decrementing prog's refcnt (through bpf_prog_put)
leaving the LSM program alive past the point of the expected lifecycle.
This fix allows for the decrement to take place.
[0] https://lore.kernel.org/all/20220628174314.1216643-4-sdf@google.com/
Fixes: 69fd337a97 ("bpf: per-cgroup lsm flavor")
Signed-off-by: Milan Landaverde <milan@mdaverde.com>
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Stanislav Fomichev <sdf@google.com>
Link: https://lore.kernel.org/r/20221213175714.31963-1-milan@mdaverde.com
BPF selftests require CONFIG_FUNCTION_ERROR_INJECTION to work. However,
CONFIG_FUNCTION_ERROR_INJECTION is no longer 'y' by default after recent
changes. As a result, we are seeing errors like the following from BPF CI:
bpf_testmod_test_read() is not modifiable
__x64_sys_setdomainname is not sleepable
__x64_sys_getpgid is not sleepable
Fix this by explicitly selecting CONFIG_FUNCTION_ERROR_INJECTION in the
selftest config.
Fixes: a4412fdd49 ("error-injection: Add prompt for function error injection")
Reported-by: Daniel Müller <deso@posteo.net>
Signed-off-by: Song Liu <song@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Daniel Müller <deso@posteo.net>
Link: https://lore.kernel.org/bpf/20221213220500.3427947-1-song@kernel.org
Kernel test robot reported bpf selftest build failure when CONFIG_SMP
is not set. The error message looks below:
>> progs/rcu_read_lock.c:256:34: error: no member named 'last_wakee' in 'struct task_struct'
last_wakee = task->real_parent->last_wakee;
~~~~~~~~~~~~~~~~~ ^
1 error generated.
When CONFIG_SMP is not set, the field 'last_wakee' is not available in struct
'task_struct'. Hence the above compilation failure. To fix the issue, let us
choose another field 'group_leader' which is available regardless of
CONFIG_SMP set or not.
Fixes: fe147956fc ("bpf/selftests: Add selftests for new task kfuncs")
Fixes: 48671232fc ("selftests/bpf: Add tests for bpf_rcu_read_lock()")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: David Vernet <void@manifault.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: David Vernet <void@manifault.com>
Link: https://lore.kernel.org/bpf/20221213012224.379581-1-yhs@fb.com
